Explatory Data Analysis for 2ForcedChoice Modal Experiments

1 Procedure for both experiments

• Platform and environment
  • The experiments were conducted online in PCIbex (PennController for Ibex). The interface switched to fullscreen at the start of the main task and exited fullscreen at the end.
  • Participants were instructed to use a desktop or laptop with Google Chrome, a keyboard, and a mouse or trackpad in a distraction-free setting.
  • Total session time was approximately 25 minutes.

• Consent and demographics
  • After an introduction screen, participants viewed and downloaded a consent form and indicated consent to proceed.
  • A brief demographics form collected age, gender, location (state and country), computer type, native language, and other languages.

• Design and counterbalancing
  • Aim: the experiments tested the interpretive force of must compared to other modal elements, or the absence of a modal.
  • Participants were randomly assigned (between subjects) to one of four answer sets: bare, have to, will, prob (a separate CSV file per set).
  • Experimental, filler, and attention-check trials were fully intermixed and randomized for each participant.
  • On critical trials, the modal from a participant’s assigned set was contrasted with must (for example, will vs. must).
  • Filler trials featured the remaining modal options among the two responses.
  • Attention-check trials consisted of unambiguous deductive contexts to verify attention.
  • Counts: experimental items = X; fillers = X; attention checks = X.
  • Exclusion criterion: participants were excluded if they failed more than 20% of attention-check trials.
  • Each experimental item appeared in one of three context conditions: abductive, deductive, or inference.
  • Context conditions were assigned using a Latin square, so each participant saw exactly one context per item.
  • On every trial, the two response options were order-randomized.

• Between-experiment differences:
  • Experiment 1 used SONA participants and included preambles that conveyed contextual information in an if-clause.
  • Experiment 2 used Prolific participants and included a shorter preamble consisting only of a well, then… phrase.
  • Experiment 3 used Prolific participants and replicated bare and haveto from Experiment 1
  • Experiment 4 used ‘got to’ constructions with “well then” and “if p ,” contrustions

• Instructions and practice
  • Participants read step-by-step instructions with two worked examples (one obvious and one less obvious).
  • Additional practice trials followed, using the same two-stage flow as the main task.

• Trial structure (main task)
  • Context presentation: participants saw a short, three-turn conversation (Speaker A – Speaker B – Speaker A) that provided the relevant context.
  • Dialogue (reading) phase: participants pressed the space bar to advance or were auto-advanced after 120 s. A Dialogue RT (ms) was recorded from dialogue onset to advance.
  • Choice phase: two candidate sentence completions appeared simultaneously. Participants selected the option they judged most appropriate; a 60 s timeout applied. An Answer RT (ms) was recorded from option onset to response.
  • After the response, the screen cleared and the next trial began.

• Breaks
  • Brief on-screen breaks were inserted at regular intervals (about every four trials). Participants resumed by pressing the space bar.

• Measures and logging
  • For each trial, the following were logged: item and condition labels (Type, Group, Item, Inference, ConditionName, Condition), the three dialogue lines (D1–D3), both options (A1–A2), the selected answer, Dialogue RT, Answer RT, and TrialNumber.
  • Practice trials were logged with the same structure and marked as practice.

• Debrief and redirect
  • Upon completion, results were submitted to the server, fullscreen ended, and participants were automatically redirected to a debrief and credit form hosted at UMD or Prolific.

2 Experiment 1

2.1 Participant Accuracy in Check Items

must vs. prob
Total N = 14
Bin	Count
[0.5,0.6)	3
[0.7,0.8)	6
[0.9,1]	5

must vs. will
Total N = 6
Bin	Count
[0.7,0.8)	1
[0.9,1]	5

haveto vs. must
Total N = 9
Bin	Count
[0.7,0.8)	2
[0.9,1]	7

bare vs. must
Total N = 11
Bin	Count
[0.9,1]	11

Click to expand Demographics

must vs. prob

Mean Age: 22.00 (18–49)

Subject	Age	Gender	Location	Computer	Language	Other language
S[51]	18	Male	Maryland, United States of America	Mac	English	Spanish (proficient), Darija (novice)
S[55]	30	Male	MD, USA	Mac	English
S[60]	18	Unlabeled	Maryland, United States	PC (Lenovo)	English	Cantonese
S[5]	19	Female	Maryland, USA	PC	English	Hindi, Urdu
S[6]	18	Female	MD, USA	Windows	English
S[9]	21	Female	Maryland, USA	Mac	English
S[12]	49	MALE	MARYLAND, USA	LINUX PC	ENGLISH	SPANISH, ITALIAN, RUSSIAN NONE FLUENTLY
S[17]	19	Female	Maryland, United States	Windows Laptop	English	Spanish
S[21]	18	Male	Maryland	Laptop	English
S[22]	18	Male	Maryland, United States	PC	English
S[26]	18	Female	Maryland, USA	Lenovo	English
S[27]	20	Female	Maryland, USA	Windows	English	Mandarin Chinese
S[31]	18	Female	College Park, Maryland	Mac	English
S[34]	24	Female	Maryland, United States	Mac	English	Korean

must vs. will

Mean Age: 21.67 (18–36)

Subject	Age	Gender	Location	Computer	Language	Other language
S[53]	18	male	maryland, usa	pc	english
S[57]	18	female	Maryland, USA	laptop	english
S[59]	18	Female	Maryland, USA	Mac	English	Spanish
S[10]	21	Male	Maryland, Prince George’s County	PC	English	Spanish
S[11]	19	Male	Maryland, US	Windows	English
S[20]	36	Male	Baltimore, MD	Mac	English	French, German

haveto vs. must

Mean Age: 19.00 (18–22)

Subject	Age	Gender	Location	Computer	Language	Other language
S[56]	18	female	Maryland, USA	lenovo thinkpad	english	chinese (beginner)
S[58]	19	Female	MD, USA	Mac	English
S[1]	18	Female	College Park, Maryland	Mac	English	Cantonese
S[13]	19	Male	College Park, MD, USA	Mac	English	Gujarati
S[15]	20	Male	College Park, MD	Mac	English	Telugu
S[16]	22	Male	MD, USA	PC	English	German
S[28]	18	Female	Maryland, USA	Mac	English	Spanish
S[30]	19	Male	Maryland, United States of America	PC	English
S[35]	18	Female	College Park, MD	Mac	English	Mandarin Chinese

bare vs. must

Mean Age: 21.27 (18–40)

Subject	Age	Gender	Location	Computer	Language	Other language
S[52]	18	male	MD, USA	Thinkpad X1 Carbon	English	Mandarin, Spanish, Japanese, Hindi
S[54]	20	Female	Maryland, United States	PC	English
S[3]	20	Male	Massachusetts	PC	English
S[7]	18	woman	MD, USA	Windows	English	Japanese
S[8]	40	Female	Maryland, USA	PC	English
S[14]	19	Female	Maryland, United States	Mac	English	Vietnamese
S[18]	26	male	Maryland, USA	PC	English	Spanish (L2 ~B2/C1)
S[19]	19	Female	Maryland, USA	PC	English	Spanish
S[24]	18	Female	Maryland, United States	Mac	English
S[25]	18	female	Maryland, USA	Dell laptop	English	Vietnamese
S[32]	18	female	MD, United States	Mac	English

2.2 Answer Choice Summary

We report target-response proportions by inference (a, d, i) and type (must, prob, have to, will, bare). Confidence intervals are Morey–Cousineau, truncated to [0,1].

Per-cell trials and items increase across blocks:

• prob: n = 10, N = 5
  
• will: n = 10, N = 5
  
• have to: n = 14, N = 7
  
• bare: n = 22, N = 11

Block: prob (n = 10; N = 5)

• a: prob 0.70 (0.42–0.98) > must 0.30 (0.02–0.58)
• d: must 0.70 (0.42–0.98) > prob 0.30 (0.02–0.58)
• i: prob 0.60 (0.17–1.00), must 0.40 (0.00–0.83)

Takeaway: Strong complementarity—prob peaks in a, must in d; i is mixed with wider CIs.

Block: will (n = 10; N = 5)

• a: must 1.00 (1.00–1.00); all others 0.
  
• d: must 0.80 (0.34–1.00); will 0.20 (0.00–0.66)
• i: must 0.80 (0.52–1.00); will 0.20 (0.00–0.48)

Takeaway: Will is at (or near) floor; must dominates, with a small will share in d/i.

Block: have to (n = 14; N = 7)

• a: must 0.93 (0.78–1.00); have to 0.07 (0.00–0.22)
• d: must 0.93 (0.78–1.00); have to 0.07 (0.00–0.22)
• i: must 0.71 (0.50–0.93); have to 0.29 (0.07–0.50)

Takeaway: Must is clearly preferred; i shows a modest but reliable share for have to.

Block: bare (n = 22; N = 11)

• a: must 0.86 (0.73–1.00); bare 0.14 (0.00–0.27)
• d: must 0.91 (0.73–1.00); bare 0.09 (0.00–0.27)
• i: must 0.86 (0.73–1.00); bare 0.14 (0.00–0.27)

Takeaway: Must overwhelmingly beats bare across all inference conditions; bare shows a small baseline.

2.3 Reading Times Summary

3 Experiment 2

3.1 Participant Accuracy in Check Items

must vs. prob
Total N = 21
Bin	Count
[0.5,0.6)	1
[0.9,1]	20

must vs. will
Total N = 21
Bin	Count
[0.7,0.8)	1
[0.9,1]	20

haveto vs. must
Total N = 21
Bin	Count
[0.9,1]	21

bare vs. must
Total N = 21
Bin	Count
[0.7,0.8)	3
[0.9,1]	18

Click to expand Demographics

must vs. prob

Mean Age: 45.86 (27–73)

Subject	Age	Gender	Location	Computer	Language	Other language
S[1]	44	male	TN, USA	HP PC	English
S[2]	49	male	Michigan, USA	PC	English
S[3]	46	male	texas, USA	PC	English
S[4]	52	male	Tennessee, USA	Mac	English
S[5]	32	Male	NC, USA	PC	English
S[6]	59	Female	Alabama	P.C.	English
S[7]	73	Female	Rhode Island	PC	English
S[8]	47	Male	NY, USA	PC	English
S[9]	41	male	MA, USA	PC	English
S[10]	48	Woman	NY, USA	PC	English
S[11]	56	male	Virginia, USA	PC	English
S[12]	50	female	California, United States	laptop	English
S[13]	64	Male	Illinois, United States	PC	English
S[14]	30	Female	Hawaii, USA	PC	English
S[15]	34	Female	New York, United States of America	PC	English
S[16]	40	Female	New York, USA	Laptop	English
S[17]	42	Male	Kentucky, USA	Windows laptop	English
S[18]	37	male	Indiana, United states	PC	English
S[19]	27	woman	US, TEXAS	WINDOWS	ENGLISH
S[20]	32	man	Tennessee, United States	PC	English	Spanish
S[21]	60	Male	North Carolina, United States	PC	English

must vs. will

Mean Age: 46.10 (26–67)

Subject	Age	Gender	Location	Computer	Language	Other language
S[1]	31	female	NJ USA	Mac	English
S[2]	52	Female	Minnesota, USA	PC	English	Beginning Spanish
S[3]	54	Male	CA, USA	PC	English
S[4]	30	Female	Kentucky, USA	PC	English
S[5]	43	female	Connecticut, USA	PC	English
S[6]	26	male	north carolina, united states	pc	english
S[7]	55	female	Tennessee, USA	Windows Laptop	English
S[8]	37	Female	Georgia, United States	PC	English
S[9]	36	female	Maryland United states	Chromebook	English
S[10]	43	Female	Texas, USA	Mac	English
S[11]	67	female	Wisconsin, USA	PC	English	Spanish
S[12]	46	male	tn, usa	PC	English
S[13]	49	Female	Massachusetts	PC	English
S[14]	56	female	california united states	pc	english
S[15]	62	male	ny, usa	pc	English
S[16]	51	Female	Virginia	Laptop	English
S[17]	28	Female	Maine, USA	PC	English
S[18]	59	male	Tennessee, USA	PC	English
S[19]	52	female	Virginia	PC	English	English
S[20]	29	Female	California United States	Laptop	English
S[21]	62	female	Washington, USA	PC	English

haveto vs. must

Mean Age: 37.24 (22–71)

Subject	Age	Gender	Location	Computer	Language	Other language
S[1]	24	Male	Florida, USA	PC	English
S[2]	45	Male	Oregon, USA	PC	English
S[3]	29	f	fl, usa	pc	english
S[4]	22	F	CO, USA	PC	ENGLISH
S[5]	32	male	Minnesota, USA	PC	English
S[6]	27	female	Georgia, USA	PC	English
S[7]	38	Female	Louisiana, United States	PC	English
S[8]	33	Woman	Maryland, USA	Mac	English
S[9]	58	Male	Georgia, USA	PC	English
S[10]	45	Male	MS, USA	MAC	English
S[11]	53	Female	United States	Mac	English	English
S[12]	35	Non-Binary AFAB	MN, USA	Mac	English
S[13]	42	woman	Tennessee, USA	PC(windows)	English
S[14]	30	Female	Indiana, USA	PC	English
S[15]	41	Female	California, USA	Mac	English
S[16]	28	Male	PA, USA	Linux	English
S[17]	47	Male	Michigan, United States	PC	English
S[18]	71	Female	Arizona, United States	PC	English
S[19]	26	Female	Illinois, USA	PC	English Language
S[20]	25	Woman	Delaware	PC	English
S[21]	31	Male	MA, USA	Windows	English

bare vs. must

Mean Age: 43.00 (21–66)

Subject	Age	Gender	Location	Computer	Language	Other language
S[1]	36	male	CA, USA	PC	English
S[2]	35	Male	AL, Geneva	Chromebook	English	No others
S[3]	38	male	Michigan, United States Of America	PC	English
S[4]	55	male	California, USA	PC	English
S[5]	51	Male	New York, United States	PC	English
S[6]	61	female	California, US	PC	English
S[7]	21	male	Georgia, United States	PC	English
S[8]	35	male	VA, USA	windows PC	English
S[9]	45	Male	Ohio, United States	PC	English
S[10]	66	Female	Florida, United States	PC	English
S[11]	44	Male	GA, USA	PC	English
S[12]	40	Female	Louisiana, United States	PC	English
S[13]	25	Male	United States	PC	English
S[14]	47	male	New York, United States	PC	English
S[15]	46	Female	Nevada, USA	PC	English
S[16]	47	Female	Mississippi, United States	PC	English
S[17]	43	Woman	SC, United States	Windows	English
S[18]	29	female	ohio, USA	PC	english
S[19]	39	Male	United States	PC	English
S[20]	45	Male	Indiana, USA	Laptop	English
S[21]	55	female	NY	PC	English	English

3.2 Answer Choice Summary

We report target-response proportions by inference (a, d, i) and type (must, prob, have to, will, bare). Confidence intervals are Morey–Cousineau, truncated to [0,1].

Per-cell trials and items by block:

• prob: n ≈ 40, N = 20
  
• will: n = 40, N = 20
  
• have to: n = 42, N = 21
  
• bare: n = 36, N = 18

Block: prob (n ≈ 40; N = 20)

• a: prob 0.538 (0.385–0.715) > must 0.462 (0.285–0.615)
• d: must 0.725 (0.581–0.869) > prob 0.275 (0.131–0.419)
• i: must 0.550 (0.399–0.701), prob 0.450 (0.299–0.601)

Takeaway: Clear complementarity—prob peaks in a, must in d; i is near 50/50.

Block: will (n = 40; N = 20)

• a: must 0.950 (0.885–1.000); will 0.050 (0.000–0.115)
• d: must 0.775 (0.631–0.919); will 0.225 (0.081–0.369)
• i: must 0.900 (0.790–1.000); will 0.100 (0.000–0.210)

Takeaway: Must dominates; will contributes small shares, largest in d.

Block: have to (n = 42; N = 21)

• a: must 0.929 (0.831–1.000); have to 0.071 (0.000–0.169)
• d: must 0.714 (0.549–0.880); have to 0.286 (0.120–0.451)
• i: must 0.905 (0.800–1.000); have to 0.095 (0.000–0.200)

Takeaway: Must is preferred overall; d shows the largest have to share.

Block: bare (n = 36; N = 18)

• a: must 0.778 (0.604–0.952); bare 0.222 (0.048–0.396)
• d: must 0.528 (0.334–0.722); bare 0.472 (0.278–0.666)
• i: must 0.694 (0.539–0.849); bare 0.306 (0.151–0.461)

Takeaway: Unlike Exp 1, bare takes a substantial share—especially in d—though must still leads overall.

3.3 Reading Times Summary

3.4 Age groups

The results does not seem to be a function of age. But there are some mismatches.

4 Experiment 3

4.1 Participant Accuracy in Check Items

haveto vs. must
Total N = 21
Bin	Count
[0.9,1]	21

bare vs. must
Total N = 21
Bin	Count
[0.7,0.8)	1
[0.9,1]	20

Click to expand Demographics

haveto vs. must

Mean Age: 48.57 (35–73)

Subject	Age	Gender	Location	Computer	Language	Other language
S[1]	37	male	California, USA	PC	english
S[2]	60	Female	Idaho, USA	PC	English
S[3]	47	MALE	BURKETVILLE, MD	PC	ENGLISH	NONE
S[4]	38	male	ma, usa	mac	english
S[5]	60	Female	CA, USA	Mac	English
S[6]	50	Female	California, USA	Mac	English
S[7]	42	female	RI, USA	PC	English
S[8]	73	Female	Kansas, USA	laptop	English
S[9]	35	male	Colorado, United States	PC	English
S[10]	49	Male	MN	Mac	English
S[11]	46	male	Washington, USA	pc	English
S[12]	42	Female	Virginia, USA	PC	English
S[13]	72	female	Ohio, United States	PC	English
S[14]	44	female	IL, USA	PC	English
S[15]	46	Male	Michigan, United States	PC	English
S[16]	52	male	CT USA	pc	English
S[17]	41	male	Indiana. United states	Laptop	English
S[18]	58	female	Michigan, United States	PC	English
S[19]	39	male	NH, USA	PC	English
S[20]	44	female	California, US	1pc	English
S[21]	45	Male	NC, USA	PC	English	French (Intermediate), Spanish (Basic)

bare vs. must

Mean Age: 48.52 (36–72)

Subject	Age	Gender	Location	Computer	Language	Other language
S[1]	49	woman	NY, USA	PC	English
S[2]	43	Female	New York, US	HP	English
S[3]	38	Female	New York, United States	PC	English
S[4]	38	Male	Arlington, USA	PC	English	English
S[5]	71	Male	Texas, United States	PC	English
S[6]	43	F	MA, USA	Mac	English
S[7]	45	Female	Maryland, USA	Laptop%2C Windows	English
S[8]	37	female	La Porte, Texas	PC	English
S[9]	47	Male	Florida, USA	PC	English
S[10]	52	female	Missouri, USA	Mac	English
S[11]	51	male	New Mexico, USA	PC	English
S[12]	44	Female	NJ, United States	PC	English
S[13]	51	Male	Nevada, USA	PC	English	0
S[14]	48	male	OH, USA	PC	English
S[15]	36	female	California, usa	pc	english
S[16]	72	Female	Louisiana, United States	PC	English
S[17]	57	woman	CA USA	Mac	English
S[18]	51	female	michigan	pc	english
S[19]	60	male	Missouri, USA	PC	American English	Patios
S[20]	39	male	RI, USA	Pc	English
S[21]	47	Male	PA, USA	MAC	English

4.2 Answer Choice Summary

We report target-response proportions by inference (a, d, i) and type (must, prob, have to, will, bare). Confidence intervals are Morey–Cousineau, truncated to [0,1].

Per-cell trials and items by block:

• have to: n = 42, N = 21
• bare: n = 40, N = 20

Block: have to (n = 42; N = 21)

• a: must 0.93 (0.86–1.00); have to 0.07 (0.00–0.15)
• d: must 0.88 (0.77–0.99); have to 0.12 (0.01–0.23)
• i: must 0.86 (0.76–0.95); have to 0.14 (0.05–0.24)

Takeaway: Must clearly dominates; have to contributes a small but non-zero share, largest in i.

Block: bare (n = 40; N = 20)

• a: must 0.80 (0.66–0.94); bare 0.20 (0.06–0.34)
• d: must 0.68 (0.51–0.85); bare 0.33 (0.16–0.50)
• i: must 0.78 (0.62–0.93); bare 0.23 (0.07–0.38)

Takeaway: Bare claims a substantial minority share—especially in d—though must remains the plurality across inference conditions.

4.3 Reading Times Summary

5 Experiment 4

5.1 Participant Accuracy in Check Items

gotta vs. must
Total N = 21
Bin	Count
[0.7,0.8)	2
[0.9,1]	19

gotta vs. must
Total N = 21
Bin	Count
[0.7,0.8)	2
[0.9,1]	19

Click to expand Demographics

gotta vs. must

Mean Age: 46.00 (35–65)

Subject	Age	Gender	Location	Computer	Language	Other language
S[1]	65	female	Arkansas, USA	Windows	English
S[2]	42	female	MO, USA	Mac	English
S[3]	51	Female	California, USA	PC	English
S[4]	62	male	florida, usa	pc	english
S[5]	43	Female	Kansas, USA	PC	English
S[6]	51	female	IL, United States	PC	English
S[7]	35	Female	OH, United States	PC	English
S[8]	50	male	Texas, USA	pc	English	Spanish
S[9]	39	female	Willis, TX	PC	English
S[10]	39	Male	Wyoming, USA	PC	English
S[11]	50	male	New Jersey, USA	PC	English
S[12]	38	female	Oklahoma, USA	PC	English
S[13]	37	Male	IL, USA	PC	English
S[14]	37	Male	New Jersey; United States	Desktop PC	English
S[15]	54	Female	Illinois, United States	Laptop	English
S[16]	48	Male	NH	PC (windows)	English
S[17]	40	Male	Maryland, United States	PC	English
S[18]	43	Male	Illinois, USA	PC	English
S[19]	50	male	California, USA	pc	English	Spanish
S[20]	37	Female	North Carolina, USA	PC	English
S[21]	55	Female	New York	PC	English

gotta vs. must

Mean Age: 50.05 (35–81)

Subject	Age	Gender	Location	Computer	Language	Other language
S[1]	41	female	PA, USA	PC	English
S[2]	38	Male	Illinois, USA	PC	English
S[3]	48	Male	Wisconsin, USA	PC	English
S[4]	59	woman	US	PC	English
S[5]	62	Female	California, USA	PC	English
S[6]	43	male	Florida, USA	PC	English
S[7]	47	Male	Georgia, United States	PC	English
S[8]	51	male	NC, Bladen	PC	English
S[9]	81	Female	Tennessee, USA	laptop	English
S[10]	43	female	Michigan, United States	PC	English
S[11]	44	Male	Michigan, United States	PC	English
S[12]	38	male	NC, USA	PC	English
S[13]	39	Male	Kansas, United States	PC	English
S[14]	55	female	Florida, USA	PC	English
S[15]	53	Woman	Ohio, United States	Mac	English
S[16]	49	male	florida, USA	PC	English
S[17]	53	Woman	GA, United States	PC Windows	English
S[18]	47	Male	Indiana, USA	PC	English
S[19]	65	female	North Carolina, USA	PC	English
S[20]	60	Female	MA, USA	PC	English
S[21]	35	female	CA, US	Mac	English

5.2 Answer Choice Summary

We compare the share of must vs gotta across inference conditions (a, d, i) under two preambles: an if-p conditional (cond) and a well-then discourse preamble (well).

Preamble: if-p (cond)

• a: must 0.974 (0.924–1.000); gotta 0.026 (0.000–0.076).
  
• d: must 0.947 (0.879–1.000); gotta 0.053 (0.000–0.121).
  
• i: must 1.000 (1.000–1.000); gotta 0.000.

Takeaway: Near-categorical preference for must under if-p across all inference conditions; gotta contributes only trace amounts in a/d and none in i.

Preamble: well-then (well)

• a: must 0.947 (0.879–1.000); gotta 0.053 (0.000–0.121).
  
• d: must 1.000 (1.000–1.000); gotta 0.000.
  
• i: must 0.921 (0.813–1.000); gotta 0.079 (0.000–0.187).

Takeaway: Must still dominates, but well-then allows a small gotta share—most notably in i—while d remains categorically must.

5.3 Reading Times Summary

6 Overall

Preamble	Exp	age	inference	Mean	Lower CI	Upper CI
prob
If p, ...	exp1	22	abductive	0.70	0.42	0.98
If p, ...	exp1	22	deductive	0.30	0.02	0.58
If p, ...	exp1	22	inductive	0.60	0.17	1.00
Well, then ...	exp2	46	abductive	0.54	0.38	0.72
Well, then ...	exp2	46	deductive	0.28	0.13	0.42
Well, then ...	exp2	46	inductive	0.45	0.30	0.60
haveto
If p, ...	exp1	19	abductive	0.07	0.00	0.22
If p, ...	exp1	19	deductive	0.07	0.00	0.22
If p, ...	exp1	19	inductive	0.29	0.07	0.50
Well, then ...	exp2	37	abductive	0.07	0.00	0.17
Well, then ...	exp2	37	deductive	0.29	0.12	0.45
Well, then ...	exp2	37	inductive	0.10	0.00	0.20
If p, ...	exp3	49	abductive	0.07	0.00	0.14
If p, ...	exp3	49	deductive	0.12	0.01	0.23
If p, ...	exp3	49	inductive	0.14	0.05	0.24
will
If p, ...	exp1	22	abductive	0.00	0.00	0.00
If p, ...	exp1	22	deductive	0.20	0.00	0.66
If p, ...	exp1	22	inductive	0.20	0.00	0.48
Well, then ...	exp2	46	abductive	0.05	0.00	0.11
Well, then ...	exp2	46	deductive	0.22	0.08	0.37
Well, then ...	exp2	46	inductive	0.10	0.00	0.21
bare
If p, ...	exp1	21	abductive	0.14	0.00	0.27
If p, ...	exp1	21	deductive	0.09	0.00	0.27
If p, ...	exp1	21	inductive	0.14	0.00	0.27
Well, then ...	exp2	43	abductive	0.22	0.05	0.40
Well, then ...	exp2	43	deductive	0.47	0.28	0.67
Well, then ...	exp2	43	inductive	0.31	0.15	0.46
If p, ...	exp3	49	abductive	0.20	0.06	0.34
If p, ...	exp3	49	deductive	0.32	0.15	0.50
If p, ...	exp3	49	inductive	0.22	0.07	0.38
gotta
If p, ...	exp4	46	abductive	0.03	0.00	0.08
If p, ...	exp4	46	deductive	0.05	0.00	0.12
If p, ...	exp4	46	inductive	0.00	0.00	0.00
Well, then ...	exp4	46	abductive	0.05	0.00	0.12
Well, then ...	exp4	46	deductive	0.00	0.00	0.00
Well, then ...	exp4	46	inductive	0.08	0.00	0.19